This weekend, the BBC's annual genealogy exhibition is set to attract more than 20,000 visitors. David Randall explains how he became obsessed with his roots – and why so many of us can't resist peering into the past

For the last 16 months or so I have been obsessed with a woman called Eliza Wright. She married at 18 and was a mother by the age of 20, but, of her life before she wed, I know nothing. Then again, I spend quite a lot of time mooning over Charlotte Collins, three of whose six children were born before she married. And it is rare that more than a few days go by without me thinking wistfully about Martha Rundle and the way she seemed to disappear completely after her son was christened. These are the dark ladies of my life: beguiling, mysterious, and, as it happens, dead for many, many years. Each of them was unknown to me until I started researching my family tree, but they tantalise me almost daily now – they and their lost details are roadblocks to further progress down my ancestral line.

I will be taking them to west London this weekend, hoping, at Olympia's Who Do You Think You Are Show, to find at least some solutions to their riddles. Here will be the biggest gathering of genealogical brains in Britain, and the most palpable manifestation of the family tree phenomenon which absorbs so many of us these days. In 2007, its first year, the show attracted 9,500 people; and this year approaching 20,000 are expected. Thousands like me will tour the stands – of subscription services such as Ancestry and FindMyPast, magazines, county family history societies, the Society of Genealogists and other learned bodies, publishers, software companies, DNA detectors – and many of us will clutch the names of long-lost relatives on scraps of paper, seeking some clue to their whereabouts on the maze of records now available.

I have been exploring my roots (and those of my wife Pam) for nearly a year and a half now. It began with a funeral, and Christmas – the two most common triggers for a sudden interest in your ancestry, according to Else Churchill, of the Society of Genealogists. The death was my aunt, the last person of my parents' generation whose brains could be picked, and the Christmas encounter was with my daughter-in-law's father and his family tree, which included the poet William Cowper. It made me wonder, probably for the first time, who lurked in my antecedents.

And so, armed with the notes my father once made setting out the last three generations, sans maiden names, I joined Ancestry and FindMyPast, and at the former began building my family tree. Here follows what I have learnt, minus all but the most entertaining crumbs from the Randall line – partly because I and my sons are the present culmination of several hundred years of unmitigated ordinariness, but mainly because other people's family trees are, I suspect, even less fascinating to outsiders than their holiday videos. (Family tree bore: "Now what was great uncle Arthur doing suddenly turning up in Kent, I hear you ask. Well, this brings me to the Maidstone wing of the family and their role in the early days of the cement industry..." You: "Goodness me, is that the time?").

Two things immediately help – an unusual surname (my wife's is Cleave, her ancestors include Wiblins, Squibs, and Jetts); and experience in research. I have a history degree, and have spent my life as a journalist, kicking the tyres of evidence (not always effectively), and learning not to trust something just because it is found online. This is especially valuable using Ancestry, which allows you to view someone else's family tree when a person on it seems to coincide with one of your ancestors. This facility is immensely useful (there are, after all, 20 million trees on their database), but the research of others is not always as reliable as one would wish. It is by no means uncommon to see on other trees people (born, say, in 1823) fathering children three years later; or folk born and christened in Wiltshire, marrying there in 1860, living in Kansas in 1861, and then back in Wiltshire three weeks later for the baptism of their first child. The same name does not mean the same person, you soon discover.

And then there is the sheer size, and frailty, of the records. Search William Brown on Ancestry, for example, and you get no fewer than 68,126 census matches in England alone. Add in likely mis-spellings and mis-transcriptions, both contemporary and recent, and the possibilities for not immediately laying hands on your quarry are considerable. I have found nine different spellings of my surname in various records, some the result of error, some because the names of illiterate ancestors were written down by others, not themselves. This, I suppose, is how the spelling of it mutated from Rundel to Randall in four generations. Two things help: searching using an asterisk, which can stand for any missing characters (Rand*l will scoop up Rand, Randall, Randell, Randoll etc); and searching for life events in specific counties.

The growth of records online is now well beyond birth, death, and marriage indexes (from 1837), and censuses (from 1841-1911). There are criminal records, ships' passenger lists, alien arrivals, army lists, school records, burials at sea, prison hulk registers and much more. County councils are beginning to wake up to the financial benefits of putting parish records online (going back to 1538, if you're lucky), Essex being a laudable pioneer. Yet for insight into your ancestors, you really need them to have been in prison, the workhouse or services, all of which recorded not only their doings and misdoings but details of their character and health. My great uncle Edward John Edgson was unknown to me two years ago, and even then was just a name – until I found his service record, learnt of his propensity to insult officers ("14 days confined to barracks"), and also, after serving as a military policeman in France, being treated for syphilis. There, in the facsimiles of these musty records, lies many an eye-opener for the unprepared direct descendant.

Wills, or at least the statement which accompanied the granting of probate, are available. The National Probate Calendar for England and Wales covers the period 1861 to 1941, has records for more than six million estates, and is a very informative pool to fish in. Some of its records, to take one example from my mother's family, give a glimpse into the circumstances of the death ("John Richard Edgson, who died suddenly on 19 December 1925 at the Nag's Head, Hillingdon..."); others, by probate coming decades after death, can suggest entertaining family disputes. And, I asked myself when turning up the probate for my wife's great-great-great uncle William Cleave, how did a chemist in a small Devon village accumulate an estate worth more than half a million pounds at today's values?

Sooner or later, if the process grips you, two things happen. First, your curiosity is piqued sufficiently to want to visit the places where your ancestors grew up. It is one thing to know several generations of your family lived in, say, the village of Chudleigh in Devon, quite another to go there, walk the streets, identify the premises where they were the local druggist or ironmonger, visit the church where they were baptised and married (half-closing your eyes and imagining the pews filled with Victorians) and tour the churchyard, crawling on hands and knees to brush away the lengthening grass and see that they had children who were born, grew ill and died between censuses and thus went unrecorded. And so it was, last summer, that we combed graveyards in South Devon, and one day sought a mid-morning cup of coffee at the Cridford Inn in Trusham. I mentioned our mission, a village history was produced, and we realised that the inn was formerly the house for the farm my wife's family occupied. We were having our drink in what used to be their parlour. (My wife was much moved by the contact with her forbears and bought flowers for several graves. Several weeks later, a distant member of her family wrote excitedly to say "someone was putting fresh bouquets on the graves!" We felt obliged to disillusion her.)

Second, you realise that census and other records can lie – or rather be lied to. That small child described in 1851 as the son of the couple in their late fifties, for example, is quite likely to be the bastard son of their daughter. That "husband and wife" may merely have been long-term co-habitees. And so, to verify what you see online and possibly solve a few riddles, you need to start sending off for copies of birth, death, and marriage certificates. (Ordered online, from the General Register Office, they cost £9.25 each.) Starting with your parents and working backwards you begin after months of romping disjointedly around the internet to at last record things systematically, a stage it took me nearly a year to reach. But that only works as far back as 1837 when registration began and so the time will come when it is off to county record offices for parish registers, poll books and the like. It is then I suppose that membership of a county family history society or the Society of Genealogists becomes not only a prospect, but worthwhile (the former full of helpful folk whose experience has been bought by heavy investment of time, the latter having those, but also the world's best genealogical library outside the US, and running remarkably cheap courses).

What then have I got after 16 months of ferreting around, mainly online, sometimes for many hours each winter weekend? Two family trees (wife's and mine) with 1,683 people (I tend to scoop up names from all branches so that I may coincide with the maximum number of other family trees), and continuous direct lines going back to 1680 in my case, and 1498 in my wife's (both subject, pre-1750s, to a lot of verification). You might think that somewhere in all these people would be someone of even vague public achievement, distinction or infamy (even if only an 18th-century sheep stealer). So far – apart from my father-in-law, who in 1943 won this country's second-highest award for gallantry – nothing. Nor, until the last two generations, is there much sign of appreciable social mobility, either upward or downward. Here for example are my direct male line's occupations from 1730: broadweaver, weaver, weaver's daughter, cordwainer (shoemaker), publisher's assistant, telegraphist, oil company sales manager, journalist. And here is my wife's from 1810: cooper, tailor, carpenter, builder's foreman, clerk of the works, nurse. Few people, I notice, record occupations on their trees. I sampled 100 recently and found only four. Is this merely not bothering to enter the information or is it a species of snobbery at work, a bashfulness about all those agricultural labourers?

Why do we do it? The rise of this curiosity about one's roots is normally explained by the influence of television's Who Do You Think You Are and the convenience of online access to records. But some say the researching of ancestors, the erection of them into a family tree, is a symptom of the decline of family – that families as we once knew them are now less reality and more heritage – and therefore need conscious preserving. Else Churchill rejects this idea, but I do think there is more than non-idle curiosity at work. Ancestor tracing is something you come to at a certain age (I'm nearly 60). You have time and money, but also, I think, especially after your parents die, a very clear sense of your own mortality. And you begin to realise that when you cease to exist, and those who knew you personally follow, any memory of you depends on your family. In the past, the close-knit, geographically immobile family did that preserving for you, incorporating you into family legend; in these far more fluid times, we have worked out that the only thing left for most of us when we're long gone is our place in a family. It will be our only context; our descendants our only chroniclers. Like Bronze and Iron Age folk before us, with their barrows and cairns, our memory is maintained only by our own tribe. Is it possible that this obsessing over a family tree, as we now do, is a form of self-perpetuation? It is, perhaps, the Me Generation immortalising itself.

Who Do You Think You Are? Live runs from 25 February at Olympia, London